Skip to content

For online shops and D2C brands

Audience studies with cluster analysis

You publish a study for your customers and shop visitors, and uxspire groups the answers with the help of machine learning. The result is a set of segments, each with its share, defining traits and a persona set card.

No extra snippet: you send studies by link, for example in your newsletter, or embed them with the same uxspire script as your surveys.

German-language templates, including “E-Commerce generisch”
23
segments per solution, no statistics background needed
2 to 5
completed responses, then the analysis starts
from 100
Audience map · 3 segments

Segments3

Each dot is one of the 364 responses from the e-commerce example study (in German), placed by how similar the answers are. Pick a segment to see what sets it apart.

Personas from study answers.

Many personas are created once in a workshop and never updated afterwards. An audience study builds them from the answers your customers actually gave.

Before

Scattered answers, no visible patterns.

The answers sit in a spreadsheet. Anyone who wants to find the patterns has to combine the fields by hand.

The first rows of the example study’s dataset
ResponseReason for visitDimensionsMaterial, careDelivery in 3 daysWaits for discountsMissing informationPrevious orders
r001Compare3353Ordered before
r002Other4333NothingLooked before
r003Specific product5352Material, feel, Delivery dateFirst visit
r004Inspiration3332Real colourOrdered before
r005Compare335Size in my roomLooked before
r006Inspiration5414NothingLooked before
r007Compare4322NothingLooked before
r008Inspiration4421Size in my roomLooked before

Plus 356 more rows with 15 answer fields each. Numbers: scale from 1 to 5. Empty cells: no answer.

With an audience study

Clear segments with size and defining traits.

The study groups the answers by response pattern. You get segments with their size, defining traits and persona set cards.

From answer to persona in four steps.

The study works with the answers you collect in uxspire. You do not need to export them to another tool.

Study responses364 completed
  • Reason for visitSingle choice · 364 answers describes segments
  • Stage of choiceSingle choice · 364 answers describes segments
  • Information needsLikert scale · 364 answers forms segments
  • ServiceLikert scale · 364 answers forms segments
  • Price and qualityLikert scale · 343 answers forms segments
  • Missing informationMultiple choice · 338 answers describes segments
  • Previous ordersSingle choice · 364 answers describes segments
  • Open answerOpen text · 82 answers describes segments
+ Create studyBlank · template · your own JSON template
Cluster detectionk = 3 · confirmed

Cluster map (3 clusters)

Choose the number of clusters

Choose the simplest production-ready segmentation whose evidence is comparable to the best option.

Detected segments
  • Segment 01

    High need for product-fit information

    143 respondents · core 37%

  • Segment 02

    Delivery and service focus

    99 respondents · core 59%

  • Segment 03

    Seeking inspiration, decision still open

    122 respondents · core 30%

Persona set cards
  • Strongest differences per segment, tested against chance
  • More and less often than among all other respondents
  • Targeting notes per persona
  • Adjust segment names after the analysis
  • Set cards as PDF for your team

How robust the result is

Only solutions with at least 10 people in the smallest segment and sufficiently stable segment statements can be confirmed; the others stay blocked, with the reason shown. 100 responses are a technical starting condition, not proof of representativeness. On the set card, a difference only counts once at least 30 people in the segment have answered the question and the gap to all other respondents is larger than chance; each share shows how many responses it is based on. If a segment differs in this way on fewer than 2 questions, the analysis shows “Segment without a reliable profile”. Methodology of audience studies in detail (in German)

Four views of the same clusters.

Comparison, Matrix, Radar and Focus are four analyses with different depth. Comparison shows the biggest differences between the clusters, Matrix every single question including open answers, Radar the pattern of each question and Focus a single segment in detail. All of them work on the same dataset.

  • Segment 01

    High need for product-fit information

    143 respondents · core 37%

  • Segment 02

    Delivery and service focus

    99 respondents · core 59%

  • Segment 03

    Seeking inspiration, decision still open

    122 respondents · core 30%

Comparison

Top numeric differences only. Open text answers are available in Matrix and Radar.

10 findings
FindingProduct fitDelivery and serviceInspirationDelta
Delivery within three working days4 · How important are these services to you when you order?364 answered26%86% ◆17%69 ppstrong
Dimensions and proportions3 · How important is this information to you before you order?364 answered87% ◆37%20%67 ppstrong
I am happy to wait for discounts.5 · How much do you agree with these statements?347 answered6%65% ◆34%59 ppstrong
Product in a room (examples of use)3 · How important is this information to you before you order?364 answered62%20%79% ◆59 ppstrong
I pay more for better workmanship.5 · How much do you agree with these statements?343 answered73% ◆17%43%56 ppstrong
Material and care3 · How important is this information to you before you order?364 answered80% ◆30%27%53 ppstrong
Advice by chat or phone4 · How important are these services to you when you order?364 answered53% ◆5%25%48 ppstrong
My decision is almost made2 · How far along are you with your choice?364 answered17%57% ◆9%48 ppstrong
Free returns4 · How important are these services to you when you order?364 answered41%87% ◆43%46 ppstrong
Reviews from other customers3 · How important is this information to you before you order?364 answered43%48%88% ◆45 ppstrong

Values per segment: share of people who chose an answer; for Likert questions the share of “somewhat important” or “very important”, or “somewhat agree” or “strongly agree”. Delta in percentage points.

  • Segment 01

    High need for product-fit information

    143 respondents · core 37%

  • Segment 02

    Delivery and service focus

    99 respondents · core 59%

  • Segment 03

    Seeking inspiration, decision still open

    122 respondents · core 30%

Matrix

All study questions in Builder order, with numeric differences and open text answers.

8 questions
Collect ideas for furnishing20%11%53%42 ppLook for a specific product32%42%16%26 ppCompare products I already have in mind37%13%16%24 ppComplete an order7%29%7%22 ppSomething else4%4%8%4 pp
  • Segment 01

    High need for product-fit information

    143 respondents · core 37%

  • Segment 02

    Delivery and service focus

    99 respondents · core 59%

  • Segment 03

    Seeking inspiration, decision still open

    122 respondents · core 30%

Questions

Options364 answered

Trait radar

3 How important is this information to you before you order?

Overlaps show shared ground; gaps show the dimensions that separate each cluster.

Cluster radar

12345

Answer values

Share of “rather important” or “very important”.

  • Segment 01

    High need for product-fit information

    143 respondents · core 37%

  • Segment 02

    Delivery and service focus

    99 respondents · core 59%

  • Segment 03

    Seeking inspiration, decision still open

    122 respondents · core 30%

Focus cluster

High need for product-fit information

39%

143 respondents in this cluster.

Respondents

143

Share

39%

Signals

8

Questions

8

Top gap

36 pp

Mode

Ghost compare

Defining traits

Lift versus other clusters for High need for product-fit information.

More likely than average

  • Dimensions and proportions87% (+36 pp)

    How important is this information to you before you order?

  • Material and care80% (+31 pp)

    How important is this information to you before you order?

  • I pay more for better workmanship.73% (+25 pp)

    How much do you agree with these statements?

  • Advice by chat or phone53% (+23 pp)

    How important are these services to you when you order?

  • Dimensions relative to my room47% (+19 pp)

    What information are you currently missing to decide?

Less likely than average

  • I am happy to wait for discounts.6% (−25 pp)

    How much do you agree with these statements?

  • Reviews from other customers43% (−16 pp)

    How important is this information to you before you order?

  • Delivery within three working days26% (−13 pp)

    How important are these services to you when you order?

High need for product-fit information on every question

  • 1 · Single choice · 364 answered

    What do you want to get done on today’s visit?

    • Compare products I already have in mind37%

    • Look for a specific product32%

    • Collect ideas for furnishing20%

    • Complete an order7%

    • Something else4%

  • 2 · Single choice · 364 answered

    How far along are you with your choice?

    • I have a few favourites59%

    • Just getting started24%

    • My decision is almost made17%

  • 3 · Likert · 364 answered

    How important is this information to you before you order?

    • Dimensions and proportions87%

    • Material and care80%

    • Product in a room (examples of use)62%

    • Reviews from other customers43%

    • Origin and sustainability36%

  • 4 · Likert · 364 answered

    How important are these services to you when you order?

    • Advice by chat or phone53%

    • Free returns41%

    • Delivery within three working days26%

  • 5 · Likert · 343 answered

    How much do you agree with these statements?

    • I pay more for better workmanship.73%

    • I am happy to wait for discounts.6%

  • 6 · Multiple choice · 338 answered

    What information are you currently missing to decide?

    • Dimensions relative to my room47%

    • Material, feel and care42%

    • Nothing, I have everything I need24%

    • What the colour looks like in real life21%

    • How returns work15%

    • When the delivery arrives12%

  • 7 · Single choice · 364 answered

    Have you ordered from us before?

    • No, but I have looked around before48%

    • No, this is my first visit31%

    • Yes, I have ordered before21%

  • 8 · Text

    Is there anything else we should know?

    cable10height10lamps10awkward6care6instructions6PDF6page5product5category4dimensions4fabric4

Rebuilt from the application’s components, using the data from the example study. The segment names were chosen by the example study’s team after the analysis. Details may differ.

The biggest differences between the clusters.

  • Strongest deviations. Shows what separates your audiences, sorted by the spread between the segments.

The complete question analysis.

  • Every question in study order. How each cluster scores on each answer option, with nothing left out.

Spotting patterns visually.

  • Similar or different? Shows for each question where clusters overlap and where they drift apart.

One cluster at the centre.

  • What makes the segment special? Where it responds more strongly and where more weakly than the average.

What your team gets

Persona set cards as PDF, all responses as CSV and the four views in the app. They help you make the case for decisions like these in your shop:

Product information
Which details does each group miss before buying: dimensions, materials, examples of use or reviews?
Service and delivery
For whom are fast delivery, free returns or advice decisive, and for whom do they hardly matter?
Offer and messaging
Which arguments belong in newsletters, campaigns and category texts when groups weigh price, quality and service differently?
Prioritisation
Which shop improvement comes first? Segment sizes and differences help you set a reasoned order.
Filter

Delivery within three working days

Delivery and service focus

99 respondents27% shareModerate confidence

Delivery and service

„Emphasise Delivery within three working days and avoid assuming Product in a room (examples of use) for this group.“

About

This persona represents Delivery and service focus. The draft is based on 99 respondents and 8 differentiating answer signals.

More likely

  • Delivery within three working days
  • I am happy to wait for discounts.
  • Free returns

Less likely

  • Product in a room (examples of use)
  • I pay more for better workmanship.
  • Advice by chat or phone

Targeting cautions

  • Avoid generic messaging on “How important is this information to you before you order?”; this group differs by −36 pp.
  • Avoid generic messaging on “How much do you agree with these statements?”; this group differs by −31 pp.

More-likely profile

  • Delivery within three w…86%
  • I am happy to wait for …65%
  • Free returns87%
  • My decision is almost m…57%

How to target this group

Lead with Delivery within three working days. Avoid over-weighting Product in a room (examples of use). Validate copy against the cluster table before sharing.

Built from 99 respondents

Persona setcard from the example study (in German), laid out as in the application: traits that occur more or less often in the segment than on average across all 364 respondents.

What a study asks

Segments are built from structured answers. That is why a study needs several questions with fixed answer formats, not just one open question at the end of checkout.

Start from the e-commerce template

The built-in template “E-Commerce generisch” contains 25 questions in six sections: personal details, context and usage, buying behaviour, attitudes, experience with the shop and a closing open question. You adapt it to your shop, drop questions you do not need and add what your decision requires. The template is in German.

Which answers count

Single and multiple choice, ratings, numbers and Likert scales feed into segmentation. Free text does not; you still read it per segment, because it shows reasons and wording.

Read the annotated questionnaire template (in German) · Audience analysis guide (in German)

6 sections · 25 questionsBuilt-in template

E-Commerce generisch

Three questions from the template as respondents see them, in German.

Multiple choice
Worauf legen Sie besonders großen Wert, wenn Sie im Internet einkaufen?
Single choice
Wie oft besuchen Sie den Onlineshop von [COMPANY_NAME]?
Rating
Fühlen Sie sich sicher beim Online-Shopping?

The template addresses respondents formally (“Sie”); you replace [COMPANY_NAME] with your shop name. You can change, delete or add any question. There are 23 German-language templates in total; you can upload your own as a JSON file.

How to collect enough responses

You survey people you can reach yourself. uxspire does not provide a survey panel; in return, your respondents know what they are talking about.

By link

A public link, for example in your newsletter, order confirmation or shipping email. A study shared by link needs no technical integration in your shop.

Embedded in your shop

Inline on selected pages, as a modal or on exit intent. This way you also reach visitors who have not bought yet. To do this, you set up uxspire once with one script per project; collection only starts once your consent management allows it (CMP integration).

With an eye on the sample

Existing customers from your newsletter answer differently from first-time visitors in the shop. Note who you reach through which channel so you can interpret the segments correctly.

Fieldwork example, not a benchmark

20,000 suitable contacts, 10% of them invited, 10% of those complete the study: 200 responses. Check these rates in advance against your own open and click rates. Our guide to customer surveys in online shops (in German) contains a complete fieldwork plan.

What a study costs

Audience studies are included in every plan. Your plan determines how many studies can run at the same time and how many responses count per month.

All plans and quotas

PlanConcurrent studiesResponses per monthPrice per month
Starter1500€50
Growth550,000€1,000
EnterprisecustomcustomOn request

Prices exclude VAT, as of 13 November 2026. Once approved, you start on the Demo plan: 14 days, one study with up to 500 responses.

Frequently asked questions about audience studies

Do I need statistics know-how to use the study?

No. The calculation runs in the product. You see segments, shares, the strongest differences and persona set cards, and you do not have to set up the method yourself. The methodology (in German) describes how the analysis works.

Does free text shape the segmentation?

No. Segments are built from structured answers such as choice questions, ratings, numbers and Likert scales. You see free-text answers per segment, in the Focus view also as a word cloud, but they do not change the segments.

Are 100 responses representative?

No. From 100 completed responses, the analysis can run. How robust the segments are depends on the sample, the number of responses and the quality notes. A segment’s set card needs more: differences are only tested from 30 responses in the segment per question. Statements about all customers or the market require a sample designed for that purpose.

How many segments will I get?

The analysis tests solutions with 2 to 5 segments. You choose a solution that meets the quality limits and is meaningful for your decision. In the example study, the two-segment solution is recommended; the team confirms three segments and documents why.

Can I survey people who have not bought?

Yes, if you can reach them: for example embedded on product pages or on exit intent. Newsletters mainly reach existing customers. There is no automatic detection of people who abandon their purchase; you control where surveys appear through pages and rules.

Which languages are supported?

Respondents can answer in German or English. The app interface is available in English and German. The built-in templates are currently in German; you write your own questions in your respondents’ language.

How is the data processed?

The application runs on Microsoft Azure in an EU region. The data processing agreement available online applies to Demo and self-service plans; enterprise contracts may use a separately agreed DPA. You decide which personal data you collect when you design the questionnaire; ask only what your decision requires.

Plan your first audience study.

Segments and persona set cards from your study answers. 23 templates, analysis in four views, export as PDF.


Persona set cards 23 templates Cluster method EU hosting